Rapid-Response Mode VLT/UVES Spectroscopy of GRB 060418

A&A 468, 83–96 (2007) Astronomy DOI: 10.1051/0004-6361:20066780 & c ESO 2007 Astrophysics

Rapid-response mode VLT/UVES spectroscopy of GRB 060418 Conclusive evidence for UV pumping from the time evolution of Fe II and Ni II excited- and metastable-level populations

P. M. Vreeswijk1,2, C. Ledoux1, A. Smette1, S. L. Ellison3, A. O. Jaunsen4,M.I.Andersen5,A.S.Fruchter6, J. P. U. Fynbo7, J. Hjorth7,A.Kaufer1, P. Møller8, P. Petitjean9,10,S.Savaglio11, and R. A. M. J. Wijers12

1 European Southern Observatory, Alonso de Córdova 3107, Casilla 19001, Santiago 19, Chile e-mail: [email protected] 2 Departamento de Astronomía, Universidad de Chile, Casilla 36-D, Santiago, Chile 3 Department of Physics and Astronomy, University of Victoria, Victoria, BC, Canada 4 Institute of Theoretical Astrophysics, University of Oslo, PO Box 1029 Blindern, 0315 Oslo, Norway 5 Astrophysikalisches Institut Potsdam, An der Sternwarte 16, 14482 Potsdam, Germany 6 Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA 7 Dark Cosmology Centre, Niels Bohr Institute, University of Copenhagen, 2100 Copenhagen, Denmark 8 European Southern Observatory, Karl-Schwarzschild-Str. 2, 85748 Garching bei München, Germany 9 Institut d’Astrophysique de Paris – UMR 7095 CNRS & Université Pierre et Marie Curie, 98bis Boulevard Arago, 75014 Paris, France 10 LERMA, Observatoire de Paris, 61 Avenue de l’Observatoire, 75014 Paris, France 11 Max-Planck-Institut für Extraterrestrische Physik, Giessenbachstrasse, 85748 Garching bei München, Germany 12 Astronomical Institute “Anton Pannekoek”, University of Amsterdam & Center for High Energy Astrophysics, Kruislaan 403, 1098 SJ Amsterdam, The Netherlands

Received 20 November 2006 / Accepted 11 March 2007

ABSTRACT

We present high-resolution spectroscopic observations of GRB 060418, obtained with VLT/UVES. These observations were triggered using the VLT Rapid-Response Mode (RRM), which allows for automated observations of transient phenomena, without any human intervention. This resulted in the first UVES exposure of GRB 060418 to be started only 10 min after the initial Swift satellite trigger. A sequence of spectra covering 330−670 nm were acquired at 11, 16, 25, 41 and 71 minutes (mid-exposure) after the trigger, with a resolving power of 7 km s−1, and a signal-to-noise ratio of 10−15. This time-series clearly shows evidence for time variability of 6 6 6 6 4 allowed transitions involving Fe II fine-structure levels ( D7/2, D5/2, D3/2,and D1/2), and metastable levels of both Fe II ( F9/2 4 4 and D7/2)andNiII(F9/2), at the host-galaxy redshift z = 1.490. This is the first report of absorption lines arising from metastable levels of Fe II and Ni II along any GRB sightline. We model the observed evolution of the level populations with three different excitation mechanisms: collisions, excitation by infra-red photons, and fluorescence following excitation by ultraviolet photons. Our data allow us to reject the collisional and IR excitation scenarios with high confidence. The UV pumping model, in which the GRB afterglow UV photons excite a cloud of atoms with a column density N, distance d, and Doppler broadening parameter b, = +0.06 = ± = ± = provides an excellent fit, with best-fit values: log N(Fe II) 14.75−0.04,logN(Ni II) 13.84 0.02, d 1.7 0.2 kpc, and b 25 ± 3kms−1. The success of our UV pumping modeling implies that no significant amount of Fe II or Ni II is present at distances smaller than ∼1.7 kpc, most likely because it is ionized by the GRB X-ray/UV flash. Because neutral hydrogen is more easily ionized than Fe II and Ni II, this minimum distance also applies to any H I present. Therefore the majority of very large H I column densities typically observed along GRB sightlines may not be located in the immediate environment of the GRB. The UV pumping fit also = − +0.8 constrains two GRB afterglow parameters: the spectral slope, β 0.5−1.0, and the total rest-frame UV flux that irradiated the cloud since the GRB trigger, constraining the magnitude of a possible UV flash. Key words. gamma rays: bursts – galaxies: abundances – galaxies: ISM – galaxies: distances and redshifts – galaxies: quasars: absorption lines

1. Introduction place in a star-forming region. As the GRB radiation ionizes the atoms in the environment, the neutral hydrogen and metal The inﬂuence of a γ-ray burst (GRB) explosion on its envi- column densities in the vicinity of the explosion are expected ronment has been predicted to manifest itself in various ways. to evolve with time (Perna & Loeb 1998; Vreeswijk et al. Strong observational evidence (Galama et al. 1998; Stanek 2001; Mirabal et al. 2002). Ultra-violet (UV) photons will not et al. 2003; Hjorth et al. 2003) indicates that at least some only photo-dissociate and photo-ionize any nearby molecular GRB progenitors are massive stars (Woosley 1993; MacFadyen hydrogen, but also quickly excite H2 at larger distances to its & Woosley 1999), and therefore the explosion is likely to take vibrationally excited metastable levels, which can be observed Based on observations collected at the European Southern in absorption (Draine & Hao 2002). Finally, dust grains can be Observatory, Chile; proposal no. 77.D-0661. destroyed up to tens of parsecs away (Waxman & Draine 2000;

Article published by EDP Sciences and available at http://www.aanda.org or http://dx.doi.org/10.1051/0004-6361:20066780 84 P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418

Fruchter et al. 2001; Draine & Hao 2002; Perna & Lazzati 2002; providing a 3 error circle localization. Observations with the Perna et al. 2003). Detection of these time-dependent processes, Swift X-Ray Telescope (XRT) resulted in a 5 position about one with timescales ranging from seconds to days in the observer’s minute later (Falcone et al. 2006b), which triggered our desk- frame, would not only provide direct information on the phys- top computer to activate a VLT-RRM request for observations ical conditions of the interstellar medium (ISM) surrounding with the Ultra-violet and Visual Echelle Spectrograph (UVES). the GRB, but would also constrain the properties of the emit- This was received by the VLT’s unit telescope Kueyen at Cerro ted GRB flux before it is attenuated by foreground absorbers Paranal at 3:08:12 UT. The on-going service mode exposure was in the host galaxy and in intervening gas clouds. In the X-ray, ended immediately, and the telescope was pointed to the XRT evidence has been found for a time-variable H I column den- location, all automatically. Several minutes later, the night as- sity (Starling et al. 2005; Campana et al. 2007), presumably tronomers Stefano Bagnulo and Stan Stefl identified the GRB af- due to the ionization of the nearby neutral gas. In the opti- terglow, aligned the UVES slit on top of it, and started the re- cal, none of these processes have been observed until recently, quested observations at 3:16 UT (i.e. 10 min after the Swift γ-ray when Dessauges-Zavadsky et al. (2006) reported a ∼3σ vari- detection). This represents the fastest spectral follow-up of any 6 1 ability detection of Fe II D7/2 λ2396 , observed at two epochs GRB by an optical facility (until the RRM VLT/UVES observa- roughly 16 h apart. Such observations are technically very chal- tions of GRB 060607, also triggered by our team, which were lenging because high-resolution spectroscopy combined with the started at a mere 7.5 min after the GRB; Ledoux et al. 2006). A rapidly decaying afterglow flux requires immediate follow-up series of exposures with increasing integration times (3, 5, 10, with 8−10 m class telescopes. 20, and 40 min, respectively) was performed with a slit width The Swift satellite, launched in November 2004, has per- of 1, yielding spectra covering the 330−670 nm wavelength mitted a revolution in rapid spectroscopic follow-up observa- range at a resolving power of R =∆λ/λ ∼ 43 000, correspond- tions, providing accurate (5) positions for the majority of GRBs ingto7kms−1 full width at half maximum. These observations within a few minutes of the GRB trigger. Numerous robotic were followed by a 80-min exposure in a different instrument imaging telescopes react impressively fast (within 10 s) to Swift configuration, but with the same slit width, extending the wave- triggers. As for spectroscopic observations, a number of target- length coverage to the red up to 950 nm. The data were reduced of-opportunity programs at most major observational facilities with a custom version of the UVES pipeline (Ballester et al. are regularly yielding follow-up observations of the GRB af- 2000), flux-calibrated using the standard response curves3 and terglow at typically an hour after the Swift alert. However, converted to a heliocentric vacuum wavelength scale. The log of most of these programs require significant human coordination the observations is shown in Table 1. between the science team and telescope personnel/observers. At the European Southern Observatory’s (ESO) Very Large Telescope (VLT; consisting of four unit telescopes of 8.2 m 3. Ground-state absorption lines from the host each), a Rapid Response Mode (RRM) has been commissioned galaxy of GRB 060418 to provide prompt follow-up of transient phenomena, such as 2 The spectra reveal four strong absorption-line systems at red- GRBs. The design of this system allows for completely au- shifts z = 0.603, 0.656, 1.107, and 1.490. In what follows, we tomatic VLT observations without any human intervention ex- focus on the highest-redshift absorption-line system at zabs = cept for the placement of the spectrograph entrance slit on the 1.490, which corresponds to the redshift of the GRB as shown GRB afterglow. The typical time delay, which is mainly caused below; the intervening systems are discussed in a separate paper − by the telescope preset and object acquisition, is 5 10 min, de- (Ellison et al. 2006). pending on the GRB location on the sky with respect to the At the redshift of the GRB host galaxy, we detect a large telescope pointing position prior to the GRB alert. The data number of metal-absorption lines, arising from transitions in- presented in this paper are the result of the first automatically- volving the ground state of various ions (see below), fine- triggered RRM activation. structure levels of Si II and Fe II, and metastable levels of Fe II This paper is organized as follows: the UVES observations and Ni II. We will discuss the excited-level lines in more de- and data reduction are described in Sect. 2, followed by Sect. 3, tail in Sect. 4; in this section we focus on the resonance lines, in which we discuss general properties of the absorption systems i.e. lines corresponding to an allowed transition from the ion at the host-galaxy redshift from the detection of resonance lines. ground state to a higher excited level. The ions from which res- In Sect. 4, we focus on the detection of variability of transitions onance lines are detected are C I, C IV, Cr II, Mn II, Si II, Si IV, originating from fine-structure levels of Fe II, and metastable Zn II, Mg I, Mg II, Ni II, Fe II, Al II, Al III, and Ca II. C II levels of Fe II and Ni II. The time evolution of the level popu- λ1334 is at the very blue edge of our spectrum, where the noise lation of these excited levels is modeled in Sect. 5. The results is dominating the signal. For Fe I, we determine an upper limit and their implications are discussed in Sect. 6, and we conclude (5σ) on its column density of log N(Fe I) < 11.48, using Fe I in Sect. 7. λ2484. A selection of lines is shown in Fig. 1. Because the resonance lines are not found to vary in time, we have combined all spectra (see Table 1) to achieve the highest signal-to-noise 2. UVES observations and data reduction ratio (S/N) possible. The combined spectrum has S/N = 16 On April 18 2006 at 3:06:08 UT the Swift Burst Alert Telescope at λ = 4000 Å (λrest = 1606 Å) and S/N = 36 at λ = 6500 Å (BAT) triggered a γ-ray burst alert (Falcone et al. 2006a), (λrest = 2610 Å). The redshift corresponding to the zero velocity has been adopted to be z = 1.49000. At this redshift, the Lyα 1 Lines arising from fine-structure levels are sometimes indicated line is located at 3027 Å, just outside the UVES spectral range. 6 6 with stars, e.g. Fe II for Fe II D7/2,FeII for Fe II D5/2, etc.; in We now wish to highlight a number of observations that this paper we will instead list the transition lower energy level term can be made from Fig. 1. The vast majority of the column and J value in order to indicate all levels that we will discuss in a consistent manner. 3 http://www.eso.org/observing/dfo/quality/UVES/qc/ 2 See http://www.eso.org/observing/p2pp/rrm.html response.html P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418 85

Table 1. Log of UVES observations.

a b c UT start epoch Fig. 2 line colour ∆T λcentral coverage exptime seeing FWHM S/N OT mag (2006 April 18) (min) (nm) (nm) (min) ()(kms−1) peak 3:16:07 1 black 11.47 390 328−452 3 1.56.96R (6 min) = 14.0 3:16:02 1 11.38 564 462−560; 568−665 3 1.17.214V (6.5min)= 15.0 3:20:17 2 red 16.61 390 328−452 5 1.56.96z (16.3min)= 14.4 3:20:12 2 16.52 564 462−560; 568−665 5 1.07.215 3:26:27 3 blue 25.13 390 328−452 10 1.56.97 3:26:17 3 24.96 564 462−560; 568−665 10 1.07.216 3:38:01 4 green 41.20 390 328−452 20 1.46.97 3:37:56 4 41.12 564 462−560; 568−665 20 0.97.217 3:59:30 5 magenta 70.99 390 328−452 40 1.46.97I (69 min) = 16.5 3:59:24 5 70.88 564 462−560; 568−665 40 0.97.217z (78 min) = 16.2 4:41:51 6 128.09 437 376−498 80 1.06.19V (100 min) = 18.8 4:41:46 6 128.00 860 670−852; 866−1043 80 0.86.521I (135 min) = 17.4 a Time of ﬂux-weighted mid-exposure since GRB trigger, assuming the light curve decay index α = −1.1. b The seeing has been estimated from the 2-D spectra. c Approximate magnitude of the optical transient around the time of our spectra, in ﬁlters V (Schady & Falcone 2006), R (Melandri et al. 2006), I (Cobb 2006) or z (Nysewander et al. 2006).

density of the low- and high-ionization species (as well as fine- there are two components in this red feature, where the reddest structure and metastable species) is located within a narrow component would have a very high Mg I over Zn II ratio. − −1 range of velocity (with a spread of 50 100 km s ), and seems The best-fit redshifts and broadening parameters for each to be contained within two or three main components. This component are listed at the top of Table 2, along with the fit small range in velocity for the low-ionization species is also ionic column densities (individual and the total of all compo- seen in GRB 051111 (Prochaska et al. 2006, 2007). Highly sat- nents). The fits are shown by the solid (red) line in Fig. 1. urated lines such as those from C IV, Si IV, Mg II, and Al II − −1 When comparing the resonance lines with the transitions from show components to the blue up to 200 km s , but these har- excited levels of Fe II and Ni II, it is apparent that although bour only a small fraction of the total column density. The line the profiles are very similar, the velocity spread of the lat- profile of the high-ionization C IV lines follows the profile of ter is smaller. The one exception is Al III, whose red compo- the low-ionization lines very well, even though the compari- + −1 ffi nent (i.e. the one near 15 km s ) is not consistent with the son is made di cult by the fact that the C IV lines are much resonance-line fit. The column densities that we find are con- stronger. This similarity is uncommon in QSO-DLAs (Wolfe & sistent with those found by Prochaska et al. (2007), with the Prochaska 2000). The main clump of the low-ionization line pro- exception of Fe II, where our value of log N(Fe II) = 15.07 ± files, though kinematically simple, is a complex mix of broad 0.08 is lower (at 1.8σ significance) than their adopted value and narrow components. Thanks to the high signal-to-noise ra- of log N(Fe II) = 15.22 ± 0.03. Prochaska et al. (2007) used tio, a weak narrow component is clearly observed in the profiles − −1 the transitions Fe II λλ2249 and 2260, which are not saturated. ofCrIIandMnIIat 20 km s . In the UVES spectra these lines fall right in the red CCD gap, For a quantitative analysis, we have simultaneously fit Voigt- leaving us with one unsaturated line, Fe II λ1611, which has a profiles (using the FITLYMAN context within MIDAS) to all lower oscillator strength. Therefore, we have more confidence in resonance lines with at least one non-saturated transition. The the determination by Prochaska et al. (2007); in the discussion atomic data required by the profile fits (vacuum wavelengths, that follows, it should be kept in mind that our total Fe II column oscillator strengths and damping coefficients) have been taken density is probably too low by about 0.15 dex. from Morton (2003). For the oscillator strengths of Ni II λλ1709, From the column densities we calculate the abundance ra- 1741, and 1751, the values from Fedchak et al. (2000) have tios of several ions with respect to iron for each component been adopted instead. We find that at least three components separately and for the total, adopting the solar values from are required to yield an adequate fit to the data. Although a 3- Lodders (2003). The resulting values are listed in Table 2. The component fit does not describe the blue side of a few high S/N ratio [Zn/Fe] is high compared to the global QSO-DLA popula- lines perfectly (see e.g. Zn II λ2026 in Fig. 1), it is the sim- tion (Ledoux et al. 2003; Vladilo 2004), and suggests a large plest model that fits the data adequately. Adding a component on dust depletion, especially in component 3 where [Zn/Fe] = the blue side would also require an additional red component to 1.0. The solar value that we find for [Mn/Fe] provides addi- compensate for the loss of the broad component on the red side; tional evidence for substantial dust depletion (Ledoux et al. the redshift of this additional red component would be very hard 2002; Herbert-Fort et al. 2006). Given these indications for to constrain. We note that the total column density derived would a large dust depletion, the actual value for [Si/Zn] may be hardly change if more components are used. The redshift z and 0.2−0.3 dex higher than observed ([Si/Zn]tot = −0.08), which the Doppler-broadening parameter b (in km s−1) for each com- would suggest an α-element overabundance, provided that zinc ponent are assumed to be the same for all ions. Only for Mg I we can be used as a proxy for iron peak elements. Although the had to tweak the redshift of the red component for a satisfactory values for [Zn/Fe] and [Mn/Fe] are high compared to those fit. The slightly higher redshift for this component is reasonably found in QSO-DLAs ([Zn/Fe]QSOs = 0−1and[Mn/Fe]QSOs = consistent with the profile of Mg I λ2852, but it is more probable −0.5−0.4), they are rather typical for the ISM of GRB host that the red Mg I component is blended. A third possibility is that galaxies, with [Zn/Fe]GRBs = 1−2and[Mn/Fe]GRBs = 0.1−0.3 86 P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418

Fig. 1. Absorption-line profiles for a variety of transitions detected at the GRB 060418 redshift. To all low-ionization species containing at least one non-saturated transition, we have performed simultaneous Voigt-profiles fits using a three-component model; the resulting fits are shown by the solid (red) line (see Table 2 for the fit results). The relative velocity of the different components are indicated by the (blue) vertical dotted lines; note that two components have very similar redshifts. It is clear that this 3-component fit does not describe some high S/N lines, such as Zn II λ2026, very well; we will come back to this in the discussion. We note, however, that the total column density would hardly change if additional components would be introduced. For comparison purposes, we also show transitions that we have not fit: (saturated) higher-ionization lines of C IV and Si IV, lines from Fe II fine-structure levels, and transitions originating from metastable levels of both Fe II and Ni II.

6 6 6 6 (Savaglio et al. 2003; Savaglio & Fall 2004; Savaglio 2006). This fine-structure levels of Fe II ( D7/2, D5/2, D3/2,and D1/2), as 4 dust depletion difference between QSO-DLAs and GRB host well as from transitions from metastable levels of Fe II ( F9/2 4 4 galaxies can be naturally explained if QSO sightlines on average and D7/2)andNiII( F9/2). See Figs. 4 and 5 for an illustration do not probe the central regions of galaxies (for which there is of the relevant energetically lower levels of Fe II and Ni II, in- growing evidence, e.g. Wolfe & Chen 2006; Ellison et al. 2005; cluding the first higher excited level, and the wavelength and Chen et al. 2005a), while GRB lines-of-sight do. spontaneous decay probability of the transitions between the levels. Back to Fig. 2: we overplot the series of five spectra, − 4. Detection and variability of transitions involving epoch 1 5 (see Table 1), with the colours black, red, blue, green and magenta, respectively. Comparison of the two panels shows excited levels of Fe II and Ni II clear evidence for variability of the excited-level lines, while the In the left panel of Fig. 2 we show the profiles of some se- strengths of the resonance lines are constant in time. To show lected resonance lines and transitions arising from all four this variability more clearly, we have constructed apparent P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418 87

Table 2. Ionic column densities and abundance ratios in the combined spectrum (see Fig. 1 for the corresponding proﬁle ﬁts).

± a Ion Lines used log N σlog N component 1 2 3 total zabs 1.49001(1) 1.48999(7) 1.49018(3) b (km s−1)23.6 ± 0.43.0 ± 0.46.5 ± 0.1 C I 1656 12.52 ± 0.18 12.29 ± 0.14 12.19 ± 0.17 12.83 ± 0.11 Mg I 2026 <12.07 13.01 ± 0.02 13.60 ± 0.02b 13.70 ± 0.02b Si II 1808 15.74 ± 0.02 14.85 ± 0.17 15.34 ± 0.05 15.92 ± 0.03 Cr II 2056, 2062, 2066 13.51 ± 0.02 12.46 ± 0.06 12.88 ± 0.03 13.63 ± 0.02 Mn II 2576, 2594, 2606 12.97 ± 0.02 11.82 ± 0.05 12.59 ± 0.02 13.14 ± 0.01 Fe II 1611 14.88 ± 0.10 14.03 ± 0.31 14.49 ± 0.11 15.07 ± 0.08 Ni II 1709, 1741, 1751 13.77 ± 0.03 12.82 ± 0.13 13.03 ± 0.08 13.88 ± 0.03 Zn II 2026, 2062 12.80 ± 0.02 11.98 ± 0.05 12.69 ± 0.02 13.09 ± 0.01 Abundance ratio [Si/Fe] 0.79 ± 0.10 0.75 ± 0.35 0.78 ± 0.12 0.78 ± 0.09 [Cr/Fe] 0.45 ± 0.10 0.25 ± 0.32 0.21 ± 0.11 0.38 ± 0.08 [Mn/Fe] 0.06 ± 0.10 −0.24 ± 0.31 0.07 ± 0.11 0.04 ± 0.08 [Ni/Fe] 0.14 ± 0.10 0.04 ± 0.34 −0.21 ± 0.14 0.06 ± 0.09 [Zn/Fe] 0.76 ± 0.10 0.79 ± 0.31 1.04 ± 0.11 0.86 ± 0.08 a The errors listed are the formal errors provided by FITLYMAN; although an error of 0.02 on an individual component is likely to be an underestimate, we consider the error on the total column density to be realistic. b The third component of Mg I is likely to be blended; see the discussion in the text. column density profiles based on pixel optical depths in compos- clipping factors and the order of the polynomial with which we ite spectra (Savage & Sembach 1991) for each of the Fe II fine- fit the continuum around each line, and rerun the Voigt profile structure levels and the Fe II and Ni II metastable levels; these fit. The maximum change that we find is 0.03 dex, which we add profiles, which have been smoothed with a boxcar of 5 pixels, to the formal error for the rest of the analysis. are shown in the right panel of Fig. 2. Prochaska et al. (2007) have performed time-resolved high- We estimate the formal significance of this variability by resolution spectroscopy of GRB 060418 as well. Their three measuring the equivalent width (EW) of the individual lines in spectra were taken around the same mid-exposure time as our −1 −1 between −30 km s and +40 km s at the various epochs, con- epoch 4, 5 and 6 spectra. They do not consider variability, servatively adding 3% of the EW to its formal error, due to the and report on the average column density for the four Fe II uncertainty in the placement of the continuum. Using the differ- fine-structure levels; they do not mention the metastable levels ent EW values over the different epochs and its mean, we calcu- of Fe II and Ni II. When comparing their average values with our late the chi-square and the corresponding probability with which epoch 4 column densities, the results are fully consistent within a constant equivalent width can be rejected. For the individ- the errors. ual lines shown in Fig. 2: Fe II λ2333, Fe II λ2607, Fe II λ2407, Comparison of the resonance-lines fit with the excited-lines Fe II λ2629, Fe II λ1702, Fe II λ2563, and Ni II λ2217, the sig- b nificances are 4.5σ,5.8σ,2.1σ,0.3σ,2.5σ,1.7σ, and 3.5σ,re- fit shows that the redshifts and parameters for the three com- spectively. Using several transitions originating from the same ponents are very similar, cf. Tables 2 and 3. When we run the excited-lines profile fit with the redshift and b parameter fixed at level (the same that have been used to construct the apparent column density profiles in Fig. 2), we find the following num- the values of the resonance lines, the resulting fit is very poor. bers: 8.7σ (Fe II 6D ), 7.4σ (Fe II 6D ), 3.2σ (Fe II 6D ), Therefore, although the fits provide similar results, the redshifts 7/2 5/2 3/2 and b parameters are significantly different. We already noted 0.5σ (Fe II 6D ), 2.2σ (Fe II 4F ), 1.7σ (Fe II 4D ), and 1/2 9/2 7/2 this difference in Sect. 3, and we will come back to this point in σ 4 2.5 (Ni II F9/2). Sect. 6. We have performed Voigt-profile fitting to the lines originating from the excited levels of Fe II and Ni II, independent from Atomic fine-structure levelsarecausedbyanenergysplit the resonance-line fit. The atomic data from Morton (2003) were due to the interaction of the total electron spin and total angu- adopted when available, and if not we have assumed the val- lar momentum of the electrons. The transitions between these ues from Kurucz (2003) (note that for Ni II we have divided levels are not allowed, i.e. they cannot proceed through an elec- the Kurucz oscillator strengths by two; see the discussion in tric dipole transition, and therefore the corresponding transition Sect. 5.2). The Voigt-profile fit results are shown in Table 3 and probabilities are low. The same is applicable to other energeti- Fig. 3; we only show the fit profiles for epoch 3 as the other cally lower levels, also called metastable levels, and their fine- epochs display very similar results, but with somewhat different structure levels. Figures 4 and 5 show the energy level diagrams, signal-to-noise ratios. Just as with the resonance lines, a satisfac- of selected levels of Fe II and Ni II. tory fit is found when using three components. In an initial fit, the These fine-structure and metastable levels can be popu- redshift and b parameter were free to vary from epoch to epoch. lated through (1) collisions between the ion and other particles As these were found to be constant with time, in a final fit they such as free electrons, (2) direct photo-excitation by infra-red were fixed to the averages over the five epochs; the z and b aver- (IR) photons (with specific wavelengths between 87−260 µm), ages and the corresponding standard deviations are listed at the and/or (3) indirectly through excitation by ultra-violet (UV) pho- top of Table 3. The column density errors listed in Table 3 are the tons, followed by fluorescence. Detection of transitions from formal errors provided by FITLYMAN. We have also estimated these energetically lower excited levels provides a powerful an error in the placement of the continuum by varying the sigma probe of the physical conditions in the interstellar medium 88 P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418

Fig. 2. The epoch 1-5 UVES spectra of GRB 060418 (see Table 1) are overplotted with the colours black, red, blue, green and magenta, respectively. In the left panel individual lines are shown, typical resonance lines on the left, and the lines arising from the excited levels of Fe II and Ni II on the right. The latter show evidence for a varying equivalent width as a function of time. To make this variability clearer, we have combined various lines that arise from the same level and constructed apparent column density proﬁles, smoothed with a boxcar of 5 pixels; these are shown in the right panel.

(Bahcall & Wolf 1968), where the quantities that can be derived Therefore, this (or any) Si II fine-structure level is not included depend on the excitation mechanism. in our analysis. Vreeswijk et al. (2004) noted the presence of transitions orig- More recently, even more exotic transitions involving fine- inating from the fine-structure level of Si II in the host galaxy structure levels of Fe II have been discovered in GRB sight- of GRB 030323; as these lines had never been clearly detected lines (Chen et al. 2005b; Penprase et al. 2006; Prochaska et al. in QSO-DLAs (we note that they had been detected in absorp- 2006; D’Elia et al. 2006). As noted by Prochaska et al. (2006), tion systems associated with the QSO, see Wampler et al. 1995; these lines had previously been detected in absorption in extreme Srianand & Petitjean 2001), this detection suggested an origin in environments such as Broad Absorption-Line (BAL) quasars the vicinity of the GRB. Assuming that collisions with electrons (Hall et al. 2002), η Carinae (Gull et al. 2005), and the disk were the dominant excitation mechanism, a volume density of of β Pictoris (Lagrange-Henri et al. 1988). For the Fe II fine- 2 4 −3 nHI = 10 −10 cm was derived (see also Savaglio & Fall 2004; structure level population along GRB sightlines it has been ar- 2 Fynbo et al. 2006). We note that Si II P3/2 λ1816 is not detected gued (Prochaska et al. 2006) that IR excitation is negligible, that (5σ upper limit: log N < 14.81) in the case of GRB 060418, collisional excitation is improbable (although not excluded), and 2 and that Si II P3/2 λ1533 (see Fig. 1) is severely blended with that indirect UV pumping probably is the dominant excitation Fe II λ2382 at the redshift of an intervening absorber, z = 0.603. mechanism. The detection of variability at the 3σ level (using P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418 89

Table 3. Column densities of the Fe II and Ni II excited levels, in the The lines arising from the metastable levels of both Fe II 4 4 4 three individual components and the total, at the various epochs (see ( F9/2 and D7/2)andNiII( F9/2) that we detect are the first lines Table 1) since the burst trigger. from metastable levels to be identified along any GRB sightline. However, these have also been previously detected in BAL epoch log N ± σa log N quasars (Hazard et al. 1987; Wampler et al. 1995) and η Carinae comp. 123total (Gull et al. 2005). zabs 1.48996(6) 1.49002(4) 1.49015(9) Our clear detection of time-variation of numerous transi- b 28.0 ± 3.55.0 ± 2.110.2 ± 1.5 tions involving all fine-structure levels of the Fe II ground 6 Fe II D7/2 λλ2333, 2383, 2389, 2612, 2626 state, and moreover transitions originating from metastable lev- 113.49 ± 0.02 13.27 ± 0.03 13.63 ± 0.02 13.96 ± 0.01 els of Fe II and Ni II, allows for a critical comparison of the 213.52 ± 0.02 13.17 ± 0.03 13.64 ± 0.02 13.96 ± 0.01 data with the three possible excitation mechanisms mentioned 313.48 ± 0.02 13.14 ± 0.03 13.62 ± 0.02 13.93 ± 0.01 above. However, independent of the mechanism at play, the de- 413.37 ± 0.02 13.13 ± 0.03 13.57 ± 0.02 13.87 ± 0.01 tection of time-variable absorption implies that the flux from the ± ± ± ± 513.23 0.03 13.06 0.03 13.41 0.02 13.73 0.01 GRB prompt emission and/or afterglow, directly or indirectly, is 6 Fe II D5/2 λλ2328, 2381, 2399, 2607, 2618 the cause of the line variability, and that the absorbing atoms are 113.09 ± 0.05 13.04 ± 0.03 13.45 ± 0.02 13.71 ± 0.02 located in the relative vicinity of the GRB explosion. 213.12 ± 0.04 13.06 ± 0.03 13.45 ± 0.02 13.72 ± 0.02 313.12 ± 0.04 13.02 ± 0.03 13.39 ± 0.02 13.68 ± 0.02 413.04 ± 0.04 12.96 ± 0.03 13.36 ± 0.02 13.63 ± 0.02 5. Modeling of the time evolution of Fe II ± ± ± ± 512.84 0.07 12.86 0.04 13.24 0.02 13.50 0.02 and Ni II excited levels 6 Fe II D3/2 λλ2338, 2359, 2407, 2411, 2614, 2621 112.60 ± 0.09 12.88 ± 0.03 13.17 ± 0.02 13.42 ± 0.02 5.1. Collisional excitation 212.63 ± 0.08 12.85 ± 0.03 13.18 ± 0.02 13.42 ± 0.02 312.59 ± 0.08 12.86 ± 0.03 13.18 ± 0.02 13.42 ± 0.02 We first consider the collisional model. Although Prochaska 412.56 ± 0.08 12.78 ± 0.03 13.13 ± 0.02 13.36 ± 0.02 et al. (2006) did not detect any change in column density of ex- 512.38 ± 0.13 12.67 ± 0.04 13.00 ± 0.02 13.23 ± 0.03 cited Fe II, they did suggest that variability of absorption-line 6 strengths would be inconsistent with a collisional origin of the Fe II D1/2 λλ2345, 2411, 2622, 2629 112.38 ± 0.15 12.51 ± 0.06 12.77 ± 0.03 13.06 ± 0.04 excitation. This is certainly expected to be the case for a medium 212.39 ± 0.14 12.42 ± 0.06 12.80 ± 0.03 13.06 ± 0.04 out of reach of the influence of the GRB afterglow flux, but close 312.65 ± 0.07 12.34 ± 0.07 12.77 ± 0.03 13.10 ± 0.03 to the GRB site one might expect the incidence of intense X-ray 412.44 ± 0.10 12.42 ± 0.05 12.79 ± 0.03 13.06 ± 0.03 and UV radiation to deposit a considerable amount of energy in 512.35 ± 0.13 12.21 ± 0.08 12.73 ± 0.03 12.97 ± 0.04 the surrounding medium through photo-ionization, causing a sit- Fe II 4F λλ1566b, 1612b, 1637b, 1702b, 2332, 2360 uation similar to photo-dissociation regions (PDRs), which can 9/2 −1 112.74 ± 0.33 12.79 ± 0.14 13.17 ± 0.06 13.42 ± 0.10 produce a shock front with typical velocities of 10−20 km s 212.99 ± 0.17 12.62 ± 0.19 13.26 ± 0.05 13.51 ± 0.07 and density enhancements. Relaxation of these high-density re- 312.91 ± 0.19 12.89 ± 0.10 13.40 ± 0.04 13.61 ± 0.05 gions might then result in a change in column density. The pro- 412.97 ± 0.15 13.09 ± 0.06 13.41 ± 0.03 13.68 ± 0.04 files shown in Fig. 1 are actually suggestive of such a shock front 512.54 ± 0.44 13.03 ± 0.07 13.47 ± 0.03 13.64 ± 0.06 situation: they show two main components with a velocity differ- 613.38 ± 0.12 12.53 ± 0.42 13.42 ± 0.07 13.73 ± 0.08 ence of 25 km s−1, and moreover, the lines caused by transitions 4 Fe II D7/2 λλ1635, 2563 from the metastable levels seem to be enclosed by the ground 112.36 ± 0.30 12.35 ± 0.15 12.78 ± 0.06 13.02 ± 0.10 state species of e.g. Cr II. Therefore, it is quite reasonable to 212.33 ± 0.30 12.09 ± 0.25 12.75 ± 0.06 12.95 ± 0.11 consider the collisional excitation scenario. 312.46 ± 0.20 12.35 ± 0.12 12.40 ± 0.12 12.88 ± 0.10 ± ± ± ± If collisions of the Fe II ions with electrons, protons or 411.35 0.80 12.41 0.10 12.49 0.09 12.77 0.10 HI atoms is the dominant excitation process, and the colli- 512.45 ± 0.20 11.54 ± 0.72 11.98 ± 0.30 12.61 ± 0.20 sional de-excitation rate exceeds the spontaneous decay rate, the 4 Ni II F9/2 λλ2166, 2217, 2223, 2316 population ratio between two levels i and j should follow the ± ± ± ± 112.62 0.11 12.59 0.06 13.03 0.03 13.27 0.03 Boltzmann distribution (see e.g. Prochaska et al. 2006): 212.83 ± 0.07 12.63 ± 0.06 13.10 ± 0.02 13.37 ± 0.03 312.85 ± 0.06 12.66 ± 0.05 13.12 ± 0.02 13.40 ± 0.02 ni g j − = Eij/kTex ± ± ± ± e (1) 412.82 0.06 12.68 0.05 13.22 0.02 13.45 0.02 n j gi 512.86 ± 0.06 12.61 ± 0.06 13.24 ± 0.02 13.46 ± 0.02 where g is the statistical weight of the level, Eij is the energy a The errors listed are the formal errors provided by FITLYMAN; al- jump between the levels, k is the Boltzmann constant, and Tex though an error of 0.02 on an individual component is likely to be an is the excitation temperature. We fit this Boltzmann distribution underestimate, we consider the error on the total column density to be to all available excited levels of Fe II at each epoch separately. realistic. For each epoch we fit two parameters: the density (in our case b These lines are also covered in the spectrum with setting 437 nm (see Table 1), resulting in the determination of column densities at a column density) and the excitation temperature. The resulting 6th epoch. Boltzmann fit is shown in Fig. 6. Even with 2 free parameters for each of the five epochs, the model is clearly not able to reproduce 2 = the observed column densities: the reduced chi-square is χν 95.7/(5−2) = 31.9. If we would have only detected the fine-structure lev- two different instruments) of one Fe II fine-structure line in the els of Fe II, the collisional excitation model fit would have spectrum of GRB 020813 was reported (Dessauges-Zavadsky been acceptable, as in Prochaska et al. (2006). The main cul- et al. 2006), which the authors claim to be supportive evidence prit for the poor fit is the increasing level population of the 4 for the UV pumping model. metastable level Fe II F9/2, which cannot be accommodated 90 P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418

Fig. 3. Absorption-line profile fits to selected transitions from excited Fe II and Ni II. The lower level of the transition, for which the column density is determined from the fits, is indicated in each panel. The fit results are listed in Table 3. We only show the fits for the third epoch spectra; the quality of the spectral fits for the other epochs are very similar, with only small differences due to slightly different signal-to-noise ratios (see Table 1). in the Boltzmann fit because all the other levels, with simi- observed variability is relatively smooth in time, the assumption 4 lar energies, are decreasing with time. Inclusion of Ni II F9/2 of LTE probably is valid. (E = 8393 cm−1) would make the fit even worse, as its energy In conclusion, the collisional model is rejected with high 4 −1 level is similar to the Fe II D7/2 level (E = 7955 cm ), while its confidence. observed column density is increasing with time (see the bottom panel of Fig. 8). Moreover, the best-fit Boltzmann model predicts 4 5.2. Radiative excitation by GRB-afterglow photons column densities for the fine-structure levels of Fe II F9/2 (e.g. the predicted column densities for its first fine-structure level, To verify if our observations can be explained by radiative exci- 4 Fe II F7/2,islogN = 13.4−13.6) that are inconsistent with the tation (by IR and/or UV photons), we now consider a model of a upper limits we obtain for this level (see Fig. 9). To preserve cloud with column density N (atoms cm−2), at a distance d (pc) clarity, we do not show these in Fig. 6. from the GRB. The afterglow flux will excite the atoms in the An implicit assumption in this collisional excitation model cloud, and we will calculate the atom level populations as a func- is that of local thermodynamic equilibrium (LTE), while the tion of time, to be compared with our observations. We will only observed variability of the absorption lines suggests that this consider excitation, and neglect ionization, which, as we will see may not be valid. However, using the PopRatio4 code (Silva & below, is fully justified. Viegas 2002), we find that if collisions is the dominant excita- We can describe the afterglow flux in the host-galaxy rest tion mechanism, the observed population ratios of the Fe II fine- frame by: structure levels require an electron volume density of at least − 2 ∼ 4 3 α −β × 10 ne 10 cm . As this is very high, while at the same time the 1.192 × 10−25 tobs λobs 1.083 10 pc 393 s 5439 Å d Frest = (2) 4 We note that PopRatio only includes collisions with electrons, ν 1 + z which are the dominant collision partners for temperatures below ap- −1 −2 −1 proximately 100 000 K; beyond this temperature the contribution of in erg s cm Hz , where we have used the V = 14.99 collisions with protons and HI atoms become significant (see Figs. 3 UVOT measurement at 393 s after the burst (Schady & Falcone and 7 of Silva & Viegas 2002). 2006), corrected for foreground extinction in the Galaxy with P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418 91

Fig. 5. Energy level diagram for selected levels of Ni II (see Fig. 4). 4 Note the very low A-value for the transition F9/2 to the ground state 4 (this is actually an electric quadrupole transition), while the level F9/2 can be very easily populated from the higher excited levels, one of them 4 o being D7/2. Therefore this level can be expected to be densely populated in the presence of a strong UV radiation ﬁeld. The ﬁne-structure 2 level of Ni II ground state, D3/2 has a relatively high probability of spontaneous decay, with a mean lifetime of 1/(5.4 × 10−2 s−1) = 19 s.

Fig. 4. Energy level diagram for selected levels of Fe II. For the lower levels we only show the levels for which we detect transitions, i.e. the where Aul, Bul and Blu are the Einstein coefficients for sponta- 4 4 fine-structure levels of the Fe II ground state, and F9/2 and D7/2.Note neous decay, stimulated emission and absorption, respectively, that for clarity reasons, we do not show the fine-structure levels of the 3 with Bul = Aulλ /2hc (all in cgs units), and Blu = Bul gu/gl latter. With the arrows, we indicate the most likely transitions between (g is the statistical weight of the energy level, with g = 2J + 1, these levels, including one higher excited level. For each transition we and where J is the total angular momentum of the electrons). show the wavelength in Å and the spontaneous decay Einstein coeffi- −1 Fν(τ0) is the incoming afterglow flux at the monochromatic fre- cient, Aul in s (which is proportional to the absorption coefficient Blu, see also below). The electric dipole allowed transitions are indicated quency corresponding to the transition energy, and modified by with a solid line, and the forbidden transitions (magnetic dipole or elec- the optical depth at line center (see Eq. (3.8) of Lequeux 2005): 6 tric quadrupole) with a dotted line. Note that to populate the level D1/2, − − = τ0 + − τ0 either four IR photons are required, or two UV photons, where the Fν(τ0) Fν(0)e S ν(1 e )(4) = / = / higher levels involved need to have J 7 2andJ 3 2. This, com- × −2 = 1.497 10 Nlλ bined with the much larger transition probabilities in the UV, makes the with τ0 b f , and where the oscillator strength f is UV pumping mechanism much more efficient than IR excitation. calculated from Aul, using: m cA g λ2 f = e ul u (5) = = 2 2 AV 0.74 (Schlegel et al. 1998), and in the absorber at z 1.1 8π qegl (which shows a clear 2175 Å extinction bump) with a Milky Way extinction curve and A = 0.25 (at z = 1.1), resulting in an ef- S ν is the source function (see Eq. (3.6) of Lequeux 2005). The V Doppler width, or broadening parameter, b, has been determined fective AV,z = 0 = 0.55 (Ellison et al. 2006). So the constant in − − − = from the line-profile fits to be 28 km s 1,5kms 1, and 10 km s 1 Eq. (2) is the would-be observed UVOT flux at z 0ifthere ff would not have been any foreground extinction. The best-fit af- for the three di erent components (see Table 3). We allow the terglow intrinsic spectral slope assuming these extinction values b value to vary with the aim to obtain the best-fit value, to be is β = −0.8 (see Ellison et al. 2006). However, as this value is compared with the above measurements. For many UV transitions the cloud that we model will be optically thick, and there- quite uncertain we will also determine a best-fit value for β in the ffi fit. The flux decay in time of our spectra is very well described fore we slice up the cloud in a su cient number of plane-parallel by a power law with index −1.1, which we adopt for α.Thevalue layers, so that each layer can be considered optically thin for a particular transition; we set the maximum allowed optical depth for the decay index determined from UBVRIz photometry data = reported in GCNs range from −1.1to−1.3 (Nysewander et al. for a layer to be τlayermax 0.05. 2006; Cobb 2006; Schady & Falcone 2006). For the calculation An important ingredient in the model fit is the adopted 10 atomic data values for the spontaneous decay coefficients Aul of the luminosity distance to the GRB, dl = 1.083 × 10 pc, we −1 −1 (or equivalently, f , see Eq. (5)), and that these are exactly the have adopted H0 = 70 km s Mpc , ΩM = 0.3, and ΩΛ = 0.7. The atom level population of an upper level u with respect to same as used in the Voigt profile fits performed to obtain the alowerlevell is given by the balance equation: observed column densities (see Tables 2 and 3). We made sure that this is the case. For Fe II we include the 20 lower en- −1 dN ergy levels in our calculations (up to E = 18 886.78 cm ), u = N B F (τ ) − N [A + B F (τ )] (3) dt l lu ν 0 u ul ul ν 0 and the A’s between all these lower levels are taken from 92 P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418

column densities, it does not affect the fit results since we use exactly the same oscillator strengths in our model. The number of Ni II higher excited levels included is 334, with a resulting number of transitions of 3136. We have written an IDL routine that incorporates the equa- tions above and the adopted Fe II and Ni II atomic data values, and calculates the level evolution of the atoms in the cloud as a function of time. This model is fit to the observations (using Craig Markwardt’s MPFIT routines6) with the following free parameters: the distance d, the total Fe II or Ni II column density N, the afterglow spectral slope β, the Doppler parameter b,andthe rest-frame time at which we start the calculations, t0. We note that this t0 does not provide any constraints on the shape of the light curve before the time that our first spectrum was taken (we simply extrapolate the light curve back to t0 assuming a decay index of α = −1.1), but it does constrain the total number of photons that arrived at the cloud since the GRB trigger. By selecting the levels that we loop through, we can either treat IR excitation and UV pumping separately, or combine the two in a consistent manner. For the UV transitions, we assume that the higher excited levels are merely a route to any lower level that the higher level can combine with, i.e. after excitation of a number of atoms in one timestep, all electrons are re-distributed immediately among all possible lower levels and no electrons stay in the higher excited level. This is justified by the very large spontaneous decay transition probabilities for nearly all higher excited energy levels. As a consistency check, we compared the results of our program with the PopRatio code (Silva & Viegas 2002), which computes the Fe II fine-structure dNu = level population assuming an equilibrium situation, i.e. dt 0 Fig. 6. The top panel shows the observed total column densities (see in Eq. (3). Using exactly the same Galactic UV background as in the last column of Table 3) for the fine-structure lines (open circles; PopRatio (converted to flux density by applying a factor of 4 π), 6 6 6 6 from top to bottom: D7/2, D5/2, D3/2,and D1/2, respectively), the and in the optically thin limit, the results are identical down to 4 first metastable level (filled triangles, F9/2), and the second metastable the 0.07% level. 4 level (filled squares, D7/2) of Fe II. Overplotted are the results of the best-fit Boltzmann model (collisional scenario): solid lines for the fine- 4 4 structure levels, dashed line for F9/2, and dashed-dotted for D7/2.The 5.2.1. IR excitation only best-fit Fe II ground state column density is shown by the dotted line, First we consider only IR photons to be exciting the atoms in while the best-fit excitation temperature (Tex) is depicted in the bottom panel. It is clear that the Boltzmann model can be rejected with high the cloud. The 20 lower energy levels of Fe II are included, and confidence. we do not consider Ni II. The resulting fit is shown in Fig. 7. We note that we have imposed a lower limit to the distance of 2 pc and we fixed the spectral slope at β = −0.8andtheb pa- Quinet et al. (1996). For the transitions between the lower and rameter at 18 km s−1; the value for the latter is unimportant as higher excited levels, we adopt the values by Morton (2003) all IR transitions are basically optically thin. For distances lower if available, and if not then we use those provided by Kurucz than 2 pc, the calculation would take too long to compute on our (2003)5. The number of Fe II higher excited levels included is workstation. The reason for this is that we adjust the program 456, with a resulting number of transitions of 4443. For Ni II we timestep in such a way that a maximum of 5% of all atoms can include the lower 17 energy levels, and take the A’s between be excited to the higher excited level of a particular transition in these from Quinet & Le Dourneuf (1996), complemented by each timestep; for a large photon flux this requires a very small those from Nussbaumer & Storey (1982). For the Ni II ground timestep, i.e. a large number of calculations. If we would allow state transitions corresponding to Ni II λ 1317 and 1370 we the distance to go under 2 pc, the fitting routine would try to 4 4 adopt the f values of Jenkins & Tripp (2006), and for Ni II move the levels F9/2 (dashed line) and D7/2 (below the lower λ 1454, 1709, 1741 and 1751 from Fedchak et al. (2000). For limit of the plotting range, peaking at log N = 11.6) up, which the other transition probabilities between the lower and higher would also cause the fine-structure levels (solid lines) to move up excited levels of Ni II we again use the value from Morton slightly. The final chi-square would be lower than for the present ≥ 2 = − = (2003) if available, and otherwise those from Kurucz (2003). 2 pc fit, which has χν(IR) 2571/(31 3) 91.8, but it would As the ratio of the f -values of the Jenkins et al. and Fedchak still provide an extremely poor fit to the observations. Moreover, et al. ground-state lines compared to the Kurucz values varies at such a short distance, most of the Fe II would be expected to from 1.87 to 2.59, and we find similar ratios between values of be ionized in the first place (e.g. Waxman & Draine 2000; Perna two Morton Ni II fine-structure lines and those of Kurucz, we & Lazzati 2002), and therefore we can safely reject IR excitation have divided all Kurucz Ni II A’s by a factor of two. We stress mechanism as the explanation for the observed level population that although this factor of two results in different inferred and evolution.

5 See http://kurucz.harvard.edu 6 See http://cow.physics.wisc.edu/˜craigm/idl/idl.html P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418 93

Fig. 7. The top panel shows the same as the top panel of Fig. 6, but now with the IR excitation model overplotted: solid lines for the fine- 4 4 structure levels, dashed line for F9/2.The D7/2 level fit column density does not even reach the lower limit of the plotting range. The model prediction for the evolution of the Fe II ground state column density is shown by the dotted line. Fig. 8. The top panel shows the same as the top panel of Fig. 6, but now with the UV pumping model overplotted: solid lines for the fine- 4 4 structure levels, dashed line for F9/2, and dashed-dotted for D7/2.The 4 The reason for the relatively low population of the bottom panel displays the observed total column densities for Ni II F9/2 metastable levels compared to the Fe II ground state fine- (filled circles), and the best-fit Ni II model. In this Ni II fit, all param- structure levels in the IR excitation case is not due to a lower eters except for Ni II column density were fixed to the best-fit values obtained from the Fe II fit. The model prediction for the evolution of the transition probability for the former; e.g. between the ground Ni II ground state column density is shown by the dotted line. All Fe II 6 = × −3 −1 state and its first fine-structure level D7/2, A 2.13 10 s , and Ni II column densities are very well described by the UV pumping while between the ground state and the second metastable level model. 4 −3 −1 D7/2, A = 4.74 × 10 s (see Fig. 4). The reason is the wavelength dependence to the third power of the Einstein absorption coefficient Blu (see below Eq. (3)): photons with a longer wave- we consider 20 lower and 456 higher excited levels of Fe II. length are much more likely to be absorbed. For the levels men- The resulting fit is shown in the top panel of Fig. 8. The best-fit tioned above, this makes the transition from the ground state to values for the fit parameters are as follows: log N(Fe II ground 4 / 3 ∼ = +0.06 = ± = − +0.8 = +12 D7/2 a factor of (7955 385) 9000 less likely, while the dif- state) 14.75−0.04, d 1.7 0.2 kpc, β 0.5−1.0, t0 74−11 s, ference in the observed column density is only a factor of 10. = ± −1 2 − = and b 25 3kms , and a chi-square of χν(UV Fe II) Had we only observed the variation of the fine-structure levels 26.2/(31 − 5) = 1.01. Next, we also model the evolution of the 4 4 4 of the ground state, and not the levels F9/2 and D7/2, we would Ni II F9/2 level, using 17 lower and 334 higher levels of Ni II. have not been able to reject the IR excitation model with such We fix all parameters in the Ni II fit to the best-fit values of the high confidence, as merely considering those levels results in an Fe II fit, except for the Ni II ground state column density. The 2 = − = excellent fit to the data, with χν(IR5levels) 11.0/(20 3) 0.65. resulting fit is shown in the bottom panel of Fig. 8. The reduced Prochaska et al. (2006) rejected the IR excitation scenario on the 2 − = − = chi-square is χν(UV Ni II) 5.6/(5 1) 1.4, and the best-fit basis that IR pumping is negligible at the distance limit set by Ni II column density is log N(Ni II ground state)=13.84 ± 0.02. the detection of Mg I in their spectra (which assumes that Mg I When also including the distance as a free parameter, we find and the excited material is at the same location, which need not log N(Ni II ground state) = 13.73 ± 0.02, and d = 1.0 ± 0.3 kpc, be the case; see also Sect. 6), combined with the observation that 2 − = − = with a chi-square of χν(UV Ni II) 0.72/(5 2) 0.24. UV pumping is dominant at any given distance from the GRB, From Figs. 4 and 5, it is straightforward to see why the 4 4 in the absence of severe extinction. Although these arguments levels Fe II F9/2 and Ni II F9/2 increase with time in the are strong, they are not as conclusive as our modeling results. UV pumping scenario. The route to these levels is rather quick: one out of every 5000 photons at 2600 Å will bring the ion to 5.2.2. UV pumping this excited level. We note that the higher excited level shown in Fig. 4 is just one out of many levels that allow popula- 4 After rejection of collisional and IR excitation, we now con- tion of the Fe II F9/2 level through absorption of a UV pho- sider the UV pumping scenario. In the UV model calculations ton, followed by spontaneous decay. Once in this level, it 94 P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418

6 determine the best-ﬁt UV pumping model Fe II D7/2 column densities for GRB 060418 at 2.1 and 9.2 h in the rest frame, using the redshift of GRB 020813, z = 1.255 (Barth et al. 2003). We obtain log N(trest = 2.1h)= 13.16 and log N(trest = 9.2h)= 12.45, corresponding to a decrease of a factor of 5.1, fully consistent with the result of Dessauges-Zavadsky et al. (2006).

6. Discussion Comparison of our UVES data with modeling of collisional and radiative excitation clearly shows that UV pumping by GRB afterglow photons is the mechanism responsible for the population of the Fe II and Ni II excited levels. However, in the UV pumping model, one would expect the ions in the ground state and the excited levels to be exactly at the same location and velocity. As we discussed in Sects. 3 and 4, this is not exactly the case. One simple way of resolving this apparent discrepancy is to invoke more components than the three that we resolve. A fourth and fifth unresolved component could be hidden in the resonance- line profile at the same velocities as the two main components of the excited levels. Then the observed velocity offset between the excited levels and the resonance lines can be explained if the observed resonance-line components are mainly due to gas that is further away from the GRB, so excitation of the ground state by UV pumping is negligible. This requirement of additional ground-state components is consistent with the rather poor fit of our 3-component Voigt-profile fit to the blue wing of a few high S/N resonance lines (see Fig. 1). The existence of additional components is fully consistent with the UV pumping fit results. According to this fit, the Fe II and Ni II ground state column den- Fig. 9. Comparison of the UV-model predicted column densities for sev- sities are 0.3−0.5 dex lower than what we measure in the spec- eral additional Fe II and Ni II excited levels, i.e. others than the ones tra, providing evidence for gas along the line of sight in the host shown in Fig. 8, with the 5σ upper limits that we obtain from the spec- ff 4 galaxy that is not a ected by the GRB, and thus this gas needs tra. The Fe II D5/2 level is actually detected at just above 5σ in the to be further away than the cloud that we modeled. Assuming second epoch. that the level ratio between the first fine-structure level and the Fe II ground state of this extra material is lower than 1/10, we = − − estimate a lower limit to its distance of d 9.5 kpc. takes 1/(9.2 × 10 5 s 1) = 3.0 h for the ion to decay to the Fe II Our successful fit of the UV pumping model to the observed ground state; this is longer than the time scale over which our evolution of the Fe II and Ni II excited levels has several inter- spectra were recorded (1 h in the rest frame), and explains why esting implications. 4 this level continues to rise in Fig. 8. Ni II F9/2 is even easier to The majority of the neutral gas closest to the GRB is populate through the absorption of UV photons, and will take a at 1.7 kpc. If there would have been neutral material much closer longer time to decay to the Ni II ground state: 37 h. Transitions in, we would not be able to reproduce the observed evolution of arising from these Fe II and Ni II metastable levels are there- the excited levels with our model. Naturally this value is not ac- fore excellent probes of the UV pumping mechanism, as they curate, mainly because of the uncertainty in the possible extinc- can be observed up to many hours after the GRB event. In fact, tion in between the GRB and the cloud. However, this extinction although we were the first to identify them, these lines should is probably not very high. From the dust depletion pattern in the also be present in the high-resolution spectra of GRBs 050730, host (see Sect. 3 and Prochaska et al. 2007; Savaglio 2006), we 051111(Prochaska et al. 2006, 2007) and 050922C (D’Elia et al. estimate the extinction to be low: AV ∼ 0.1. This value is in 2005). agreement with the host-galaxy extinction estimate of AV ∼ 0.2 As we calculate the level population of the lower 20 lev- of Ellison et al. (2006), which is obtained from modeling the els for Fe II, and lower 17 levels for Ni II, we can compare if spectral slope. Moreover, the spectral slope from the UV pump- the model predictions for all levels are consistent with our data. = − +0.8 ff ing fit, β 0.5−1.0 is not very di erent from the observed spec- Searching for the detections of lines originating from these levels tral slope after correcting for the extinction in the foreground 4 has resulted in one new detection (epoch 2 for level Fe II D5/2), absorber at z = 1.1(β = −0.8), while a large amount of dust but the rest we can only place an upper limit (we adopt 5σ)to extinction would severely affect the value of the spectral slope, the column density, as shown in Fig. 9. provided that the extinction is not grey. Dessauges-Zavadsky et al. (2006) have reported a signifi- We note that another lower limit to the absorber distance is cant (∼3σ) decline by at least a factor of five in the equivalent set by the presence of Mg I (see Prochaska et al. 2006), assuming 6 width of the Fe II D7/2 λ2396 transition, in spectra taken at 4.7 that it is at the same location as a large part of the Fe II and and 20.8 h after GRB 020813. We note that this line is saturated Ni II excited material. For GRB 051111, Prochaska et al. (2006) 6 in our spectra, and moreover blended with Fe II D5/2 λ2396, and calculate that if Mg I would be at a distance smaller than 80 pc therefore we do not use it in our analysis. To verify if this decline from the GRB, then Mg I would have been fully ionized. Using is roughly consistent with our calculations for GRB 060418, we Eq. (2) of Prochaska et al. (2006), we estimate the lower limit P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418 95 to the distance where Mg I can survive to be d = 45 pc for such a gradual ionization; very likely observations have to be GRB 060418. More recently, Chen et al. (2006) have estimated performed even quicker than the response time of our observa- a lower limit to the distance from GRB 021004 to the absorbers tions, to be able see this effect. along its sightline that are blue-shifted by 2500−3000 km s−1. Our UV pumping fit shows that it is possible to obtain the From the ratio of C II/C II they find d > 1.7 kpc; limits for distance of the excited absorbing gas to the GRB. We have mod- the other absorbers blue-shifted by less than 2000 km s−1 are not eled the observations with only one cloud, but this can be ex- given. A similar distance limit is set for the Si II gas associated tended to a multiple cloud model, where for each cloud one can with GRB 050730 (Chen et al. 2006). All these distance limit obtain the distance with respect to the GRB, its velocity, its Fe II estimates are fully consistent with our distance determination. and Ni II column density and the cloud Doppler parameter, b. The consequence is that any pre-GRB neutral cloud that was And for components not affected by the UV photons, a lower present at distances less than about 1.7 kpc, was severely affected limit to the distance can be set. This way it could be possible to by GRB 060418. Atomic species typical of the neutral ISM such study the host-galaxy cloud structure, abundances and kinemat- as Fe II, Cr II, Zn II, etc., are likely ionized to a higher ion- ics in more detail than before. ization stage. We note, however, that this does not imply that Finally, the UV pumping fit not only constrains the proper- the GRB has ionized all neutral material up to this distance; ties of clouds in the GRB host galaxy, but also two properties it may be that the GRB ionized only its immediate surround- of the GRB emission. One is the spectral slope of the GRB af- ings, e.g. up to tens of parsecs, and that between the ionized terglow, even though this value is not very tightly constrained in = − +0.8 region and 1.7 kpc, no significant amount of neutral material our fit: β 0.5−1.0. The second property is the total UV flux that was present. But it is clear that the immediate environment of arrived at the cloud from the time of the burst trigger, i.e. opti- GRBs cannot be probed with these neutral ISM species, but pos- cal flash (if present) and afterglow combined. This flux can be sibly higher ionization lines may be detected. It is therefore of derived by determining the integral from fit parameter t0 to any great interest to look out for higher ionization species not nor- time desired of the assumed light curve in the model. So even mally seen in optical spectra, that may originate from the im- if no UV/optical observations were performed by robotic tele- mediate surroundings of the GRB. Possibly an ionization strat- scopes or Swift itself, the magnitude of a UV flash can be con- ification could be observed, from higher ionization lines close strained from a UV pumping fit. For the case of GRB 060418, we to the GRB, to lower ionization species further out. X-ray spec- determine the limit on the total observer’s frame V-band flux that troscopy instead of the optical will probably be the best tool to arrived from the GRB (UV flash and afterglow) from the time of probe the immediate vicinity of the GRB. the GRB trigger to the start of our first spectrum (t = 11.4min) − − − Because the photon energy threshold to ionize Fe II to to be (1.9±0.3)×10 23 erg cm 2 Hz 1. For comparison, this flux Fe III is higher than the ionization potential of H I, this dis- is the same as that contained by a V = 10 flash with a duration tance limit also applies to neutral hydrogen, i.e. any significant of 5 s. H I cloud that was present before the GRB exploded at distances smaller than approximately 1.7 kpc, will have been ionized. If we assume that GRB 060418 is not special with respect 7. Conclusions to other GRBs in this respect, most of the high-column density Using the VLT Rapid-Response Mode in combination with H I clouds observed in GRB afterglow spectra (Vreeswijk et al. UVES, we have obtained a unique time-series of high-resolution 2004; Jakobsson et al. 2006) may also be at typical kiloparsec spectra of GRB 060418 with a signal-to-noise ratio of 10−20. distances. This assumption that GRB 060418 is not special, is These spectra show clear evidence for variability of transi- supported by the lower limit to the distances of absorbers along tions arising from the fine-structure levels of Fe II, and from the GRBs 021004 and 050730 sightlines determined by Chen metastable levels of both Fe II and Ni II. We model the time evo- et al. (2006), which are also of kiloparsec scale. These H I clouds lution of the Fe II and Ni II excited levels with three possible ex- could either be part of a giant star-forming region in which the citation mechanisms: collisions, excitation by IR photons only, GRB was born, or simply clouds in the foreground in the host and UV pumping. We find that the collisional and IR photon sce- galaxy. We note that if the H I clouds are indeed at kiloparsecs narios can be rejected. Instead, the UV pumping model, in which from the GRB, any metallicity estimate performed using optical a cloud with total column density N and broadening parameter b spectroscopy is most likely not representative of the metallicity at a distance d from the GRB is irradiated by the afterglow pho- of the region where the GRB progenitor was born. The kilopar- tons, provides an excellent description of the data. The best-fit sec distance for these absorbers, combined with the significant values are log N(Fe II) = 14.75+0.06,logN(Ni II) = 13.84 ± 0.02, ff −0.04 di erences between GRB-DLAs and QSO-DLAs in H I column = ± = ± −1 density (Vreeswijk et al. 2004; Jakobsson et al. 2006), metallic- d 1.7 0.2 kpc, and b 25 3kms . The main consequence ity (Fynbo et al. 2006) and dust depletion (Savaglio 2006, see of this successful fit, is the absence of neutral gas, in the form of also Sect. 3), suggests that QSO sightlines are not probing the low-ionization metals or H I, at distances shorter than 1.7 kpc. central kiloparsecs of (GRB) star-forming galaxies. This is con- Any pre-explosion neutral cloud closer to the GRB must have sistent with our observation in Sect. 3, that the high-ionization been ionized by the GRB. Therefore, the majority of very large profiles follow those of the low-ionization species very well in H I column densities typically observed along GRB sightlines GRB 060418, while this is uncommon in QSO-DLAs. may not be in the immediate surroundings of the GRB; they could either be part of a large star-forming region, or foreground The gradual ionization of the neutral hydrogen close to the material in the GRB host galaxy. In either case, the metallicity GRB could be observed by monitoring the evolution of the Lyα derived from absorption-line spectroscopy may not be represen- or metal line (Perna & Loeb 1998, we note that these authors tative of the metallicity of the region where the GRB progenitor suggest Mg II as metal line probe, but this line is normally highly was born. saturated, especially in the high density GRB host galaxy environments, and therefore not suitable for this purpose). The red- Acknowledgements. We are very grateful for the excellent support of the shift of GRB 060418 is too low for Lyα to be covered by our Paranal staff, and in particular that of Stefano Bagnulo, Nancy Ageorges UVES spectra. From the metal lines, we do not see any hint for and Stan Stefl. We have made extensive use of the atomic spectra 96 P. M. Vreeswijk et al.: Rapid-response Mode VLT/UVES spectroscopy of GRB 060418 database of the National Institute of Standards and Technology (NIST), Lagrange-Henri, A. M., Vidal-Madjar, A., & Ferlet, R. 1988, A&A, 190, 275 see http://physics.nist.gov/PhysRefData/ASD/index.html.Thisre- Ledoux, C., Bergeron, J., & Petitjean, P. 2002, A&A, 385, 802 search was supported by NWO grant 639.043.302 to RW. This work has benefit- Ledoux, C., Petitjean, P., & Srianand, R. 2003, MNRAS, 346, 209 ted from collaboration within the EU FP5 Research Training Network “Gamma- Ledoux, C., Vreeswijk, P., Smette, A., Jaunsen, A., & Kaufer, A. 2006, GCN, Ray Bursts: an Enigma and a Tool”. 5237 Lequeux, J. 2005, The interstellar medium (Translation from the French lan- guage edition of: Le Milieu Interstellaire, ed. J. Lequeux, EDP Sciences, References 2003), Astronomy and astrophysics library (Berlin: Springer) Lodders, K. 2003, ApJ, 591, 1220 Bahcall, J. N., & Wolf, R. A. 1968, ApJ, 152, 701 MacFadyen, A. I., & Woosley, S. E. 1999, ApJ, 524, 262 Ballester, P., Modigliani, A., Boutquin, O., et al. 2000, Messenger, 101, 31 Melandri, A., Gomboc, A., Mundell, C. G., et al. 2006, GCN, 4968 Barth, A. J., Sari, R., Cohen, M. H., et al. 2003, ApJ, 584, L47 Mirabal, N., Halpern, J. P., Kulkarni, S. R., et al. 2002, ApJ, 578, 818 Campana, S., Lazzati, D., Ripamonti, E., et al. 2007, ApJ, 654, L17 Morton, D. C. 2003, ApJS, 149, 205 Chen, H.-W., Kennicutt, Jr., R. C., & Rauch, M. 2005a, ApJ, 620, 703 Nussbaumer, H., & Storey, P. J. 1982, A&A, 110, 295 Chen, H.-W., Prochaska, J. X., Bloom, J. S., & Thompson, I. B. 2005b, ApJ, Nysewander, M., Ivarsen, K., Foster, A., et al. 2006, GCN, 4971 634, L25 Penprase, B. E., Berger, E., Fox, D. B., et al. 2006, ApJ, 646, 358 Chen, H.-W., Prochaska, J. X., Bloom, J. S., et al. 2006, ApJ Perna, R., & Lazzati, D. 2002, ApJ, 580, 261 [arXiv:astro-ph/0611079] Perna, R., Lazzati, D., & Fiore, F. 2003, ApJ, 585, 775 Cobb, B. E. 2006, GCN, 4972 Perna, R., & Loeb, A. 1998, ApJ, 501, 467 D’Elia, V., Piranomonte, S., Fiore, F., et al. 2005, 4044 Prochaska, J. X., Chen, H.-W., & Bloom, J. S. 2006, ApJ, 648, 95 D’Elia, V., Fiore, F., Meurs, E., et al. 2006, A&A Prochaska, J. X., Chen, H.-W., Bloom, J. S., et al. 2007, ApJS, 168, 231 Dessauges-Zavadsky, M., Chen, H.-W., Prochaska, J. X., Bloom, J. S., & Barth, Quinet, P., & Le Dourneuf, M. 1996, A&AS, 119, 99 A. J. 2006, ApJ, 648, L89 Quinet, P., Le Dourneuf, M., & Zeippen, C. J. 1996, A&AS, 120, 361 Draine, B. T., & Hao, L. 2002, ApJ, 569, 780 Savage, B. D., & Sembach, K. R. 1991, ApJ, 379, 245 Ellison, S. L., Kewley, L. J., & Mallén-Ornelas, G. 2005, MNRAS, 357, 354 Savaglio, S. 2006, New Journal of Physics, 8, 195 Ellison, S. L., Vreeswijk, P., Ledoux, C., et al. 2006, MNRAS, L82 Savaglio, S., & Fall, S. M. 2004, ApJ, 614, 293 Falcone, A. D., Barthelmy, S. D., Burrows, D. N., et al. 2006a, GCN, 4966, 1 Savaglio, S., Fall, S. M., & Fiore, F. 2003, ApJ, 585, 638 Falcone, A. D., Burrows, D. N., & Kennea, J. 2006b, GCN, 4973, 1 Schady, P., & Falcone, A. D. 2006, GCN, 4978 Fedchak, J. A., Wiese, L. M., & Lawler, J. E. 2000, ApJ, 538, 773 Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525 Fruchter, A., Krolik, J. H., & Rhoads, J. E. 2001, ApJ, 563, 597 Silva, A. I., & Viegas, S. M. 2002, MNRAS, 329, 135 Fynbo, J. P. U., Starling, R. L. C., Ledoux, C., et al. 2006, A&A, 451, L47 Srianand, R., & Petitjean, P. 2001, A&A, 373, 816 Galama, T. J., Vreeswijk, P. M., Van Paradijs, J., et al. 1998, Nature, 395, 670 Stanek, K. Z., Matheson, T., Garnavich, P. M., et al. 2003, ApJ, 591, L17 Gull, T. R., Vieira, G., Bruhweiler, F., et al. 2005, ApJ, 620, 442 Starling, R. L. C., Vreeswijk, P. M., Ellison, S. L., et al. 2005, A&A, 442, L21 Hall, P. B., Anderson, S. F., Strauss, M. A., et al. 2002, ApJS, 141, 267 Vladilo, G. 2004, A&A, 421, 479 Hazard, C., McMahon, R. G., Webb, J. K., & Morton, D. C. 1987, ApJ, 323, 263 Vreeswijk, P. M., Ellison, S. L., Ledoux, C., et al. 2004, A&A, 419, 927 Herbert-Fort, S., Prochaska, J. X., Dessauges-Zavadsky, M., et al. 2006, PASP, Vreeswijk, P. M., Fruchter, A., Kaper, L., et al. 2001, ApJ, 546, 672 118, 1077 Wampler, E. J., Chugai, N. N., & Petitjean, P. 1995, ApJ, 443, 586 Hjorth, J., Sollerman, J., Møller, P., et al. 2003, Nature, 423, 847 Waxman, E., & Draine, B. T. 2000, ApJ, 537, 796 Jakobsson, P., Fynbo, J. P. U., Ledoux, C., et al. 2006, A&A, 460, L13 Wolfe, A. M., & Prochaska, J. X. 2000, ApJ, 545, 591 Jenkins, E. B., & Tripp, T. M. 2006, ApJ, 637, 548 Wolfe, A. M., & Chen, H.-W. 2006, ApJ, 652, 981 Kurucz, R. L. 2003, in IAU Symposium, ed. N. Piskunov, W. W. Weiss, & D. F. Woosley, S. E. 1993, ApJ, 405, 273 Gray, 45